This project seems interesting, if it means an ARM-based USB interface chip isn't needed:
Yup it's great work by Luke, I prefer the co-pro approach though during development though as it's much faster for me. It programs the FPGA with the bitmap directly whereas this approach has 2 stages; Program the flash first and then restart the FPGA and have it load the bitmap from the flash (twice as long effectively).
... but if the co-processor was a Risc-V FPGA implementation, rather than an Arm chip, using something like Luke's USB implementation (e.g like a Fomu), then it would be open source all the way down.
Yes this is an interesting idea and one which I continue to explore. In the short term however, we need mixed signal support and currently ARM derivatives provide the best bang for buck in this area, hence the current strategy.
Yes, unfortunately, you would need to replace the ARM chip with an FPGA chip, an SPI flash chip, and ADC chips which would add a lot to the cost.